Announcement: KUPAC 9th Physical AI Study Group

Kenshiro Minamikawa

M2, Graduate School of Informatics, Kyoto University / KUPAC Co-Vice Representative

「Quadruped Robot Motion Generation and the KUPAC Quadruped Robot Project」

In recent years, dynamic whole-body motion by humanoid robots has attracted increasing attention, and quadruped robots have long been one of the research areas supporting its foundations. A wide range of motion generation and control methods have been studied, from stable locomotion to rough-terrain traversal, stair climbing, jumping, and parkour. In this talk, we will review how quadruped robot motion generation has evolved, from model-based control to reinforcement learning, imitation learning, and motion generation using environmental perception. We will then organize the current state of the field and remaining challenges, discuss how motion generation may develop in the future, and introduce the quadruped robot project that KUPAC plans to pursue.

Fumiya Matsuzaki

D3, Graduate School of Mathematics for Innovation, Kyushu University

「Robot Control That Considers Intervals Between Sample Points」

Robots move continuously, but sensor measurements and control input updates are performed at fixed time intervals. As a result, even if the robot is moving correctly at the measured instants, oscillation or deviation may occur between those points. This talk focuses on motion between measurements and introduces Lifted NMPC, a control method for moving robots more accurately.